In many optical systems, such as pumps in optical communication systems, fiber gyros, control channels in optical amplifiers, sensors, and inteferometers, it is desirable to have a depolarized source to reduce polarization effects. In particular, Raman amplifiers are becoming increasingly used for distributed or remote amplification. These amplifiers are useful because they may be implemented in the normal fiber that carries the optical communications signal, and as a result, systems may be designed where amplification occurs in the transmission link itself, allowing greater transmission distances between amplifiers, higher receiver sensitivities, and lower transmission powers. Lower transmission powers have the added advantage of reducing nonlinear effects in the fiber. Furthermore, the wavelength at which the fiber Raman amplifier operates is determined by the wavelength of the pump light, unlike a rare earth-doped fiber amplifier whose gain bandwidth is limited by the rare earth species doped in the fiber. Raman is also a very low noise process, limited primarily at low powers by pump noise and quantum noise effects.
One particular concern with Raman amplification is the polarization dependence of the amplification process. The Raman gain coefficient when the pump light is polarized parallel to the signal polarization is about an order of magnitude greater than when the pump polarization is orthogonal to the signal polarization.
A long-haul single mode optical fiber is usually non-polarization preserving due its lower cost and superior performance compared to polarization maintaining fiber. Therefore, the polarizations of the pump and signal to change relative to each other as they propagate along the length of the fiber amplifier. In practice, Raman amplifiers are long, up to several kilometers in length, which permits sufficient variation in the relative polarizations of the signal and the pump that the polarization-dependent gain (PDG) effect is averaged out.
However, there remains a possibility that the pump and signal maintain the same relative polarizations for a significant length of the fiber amplifier. In such situations, the amplifier gain may be strongly dependent on the relative polarizations of the signal and the pump. This leads to uncertainty in amplifier performance, which results in increased errors in signal detection or increased system margin requirements.
One approach to reducing the effect of PDG is to use pump source light that is “unpolarized,” or that has a low degree of polarization. Unpolarized light is light that does not have a dominant polarization at an instantaneous point in time, or for which the dominant polarization changes more quickly than an amplifier response time.
Unpolarized light is further characterized as light which, when passed through a linear or circular polarization analyzer and incident on an optical power meter, shows no appreciable variation in transmitted power with analyzer azimuthal direction, and has no phase relationship between the two orthogonal polarization states.
As understood in this application, light is considered depolarized if the intensity of the light output from the depolarizer is substantially equally divided between orthogonal polarization states with no phase correlation between these two components. The use of unpolarized light to pump an optical gain medium alleviates the PDG effects, making it desirable for many optical pumping applications.
In the past, different techniques have been used to depolarize the light from a polarized light source. One conventional method for converting polarized light to depolarized light is to launch the polarized beam into a single piece of polarization maintaining (PM) optical fiber so that the launch angle is substantially at an angle of 45° relative to the birefringence axes of the PM fiber. Such a device is called a single-stage fiber Lyot depolarizer. Certain variations on this method are shown in U.S. Pat. No. 5,692,082 by N. Fukushima of Fujitsu Limited. The length of the PM fiber, as taught by Fukushima, is set so that an optical path length difference for the two polarization modes propagating through the PM fiber is greater than the coherence length of the incident light. As such, the two polarization modes are phase decorrelated, and the polarization state of the light output from the fiber is effectively randomized. However, the depolarizer in accordance with the teaching of Fukushima is costly because it requires a very long piece of PM fiber (e.g., ˜100 meters for a longitudinal mode linewidth of the order of 10 GHz). In addition it has proved highly temperature sensitive. Indeed it was observed that the degree of polarization (DoP) fluctuates rapidly and with a large amplitude (˜15-20%) over a temperature range as small as a few degrees. Thus this approach is only suitable for pump applications that can tolerate a DoP of approximately 20%.
In an article entitled “Degree of Polarization in the Lyot Depolarizer”, published in the Journal of Lightwave Technology, Vol. LT-1, No. 3, September 1983, William K. Burn teaches the calculation of optimum birefringent fiber lengths for a two-stage Lyot depolarizer for depolarizing the output of a laser source having a single longitudinal mode. In the case of a single longitudinal mode, the coherence function of the spectrum decreases monotonically with increased fiber length. Thus, as long as a minimum depolarizing fiber length is provided, a precise length is not necessary.
More recently, Hiroyuki Koshi of The Furukawa Electric Co. Ltd discloses in U.S. published application Ser. No. 0,025,111 A1 published Feb. 28, 2002, the use of a fiber depolarizers with Fabry-Perot lasers in a multiple pump architecture. Each depolarizer is a two-stage fiber Lyot depolarizer the lengths L1:L2 having a ratio of 1:2. The lengths of the PM fiber are adjusted in accordance with the coherence length of the incident light. Koshi states that “the effect of canceling the degree of polarization of the pumping source does not change even if the lengths of the polarization-maintaining optical fibers are set longer than the optimal lengths.” The Fabry-Perot lasers used in the Koshi device produce a multiple longitudinal mode spectral output, not a single longitudinal mode as assumed by Burn. As shown by the present invention, the DoP can actually be made worse by extending the length of the PM fibers beyond the optimum length. Koshi did not realize this. Accordingly, the temperature instability exhibited in the prior art is not adequately corrected in the Koshi disclosure.
A low cost, low loss temperature insensitive depolarizer is highly desired for Raman pumping architectures based on high power laser diodes, particularly a depolarizer capable of providing a DoP of less than 10%.